░  ░░░░  ░░        ░░        ░░  ░░░░  ░░░░░░░░        ░░  ░░░░  ░░       ░░░        ░░        ░
▒  ▒  ▒  ▒▒  ▒▒▒▒▒▒▒▒  ▒▒▒▒▒▒▒▒  ▒▒▒  ▒▒▒▒▒▒▒▒▒▒▒▒  ▒▒▒▒▒  ▒▒▒▒  ▒▒  ▒▒▒▒  ▒▒  ▒▒▒▒▒▒▒▒  ▒▒▒▒▒▒▒
▓        ▓▓      ▓▓▓▓      ▓▓▓▓     ▓▓▓▓▓▓▓▓▓▓▓▓▓▓  ▓▓▓▓▓        ▓▓       ▓▓▓      ▓▓▓▓      ▓▓▓
█   ██   ██  ████████  ████████  ███  ████████████  █████  ████  ██  ███  ███  ████████  ███████
█  ████  ██        ██        ██  ████  ███████████  █████  ████  ██  ████  ██        ██        █
                                                                                                

SECOND-ORDER LLM SYSTEM

PLANNING

I really wasn't entirely sure what I wanted to do for this week. I don't know much about AI prompting so it took me a bit of peripheral research to grasp what this week was asking of me. I do have experience with programming in a linux terminal environment from a C and linux class I took though. Eventually I settled on doing something pretty simple for a lack of more creative ideas. I wanted to make a nice looking website, maybe using three.js, that'd change appearance depending on the real-time weather on campus (Loop). I started by making a git repo in the folder I made and connected it to my Github account in case the server had to be reset.

After this point, I wasn't entirely sure where to go next. I had to consult ChatGPT on what to do. I was going to use the server instance of Claude for the website itself, but wanted a set of instructions for how to set up my environment first. Despite being fairly familiar with terminal my experience was mostly in programming games and compiling in C, so hosting a website and pulling from API is new to me.

For the API platform, I went with using OpenWeatherMap. I set up my file system the LLM instructed me to and put the API key in a hidden file ignored by git. I also got a basic vibecoded three.js scene working with some rectangles to represent buildings, and lighting based on weather and time. This is sort of where I lost my footing though. Because when trying to set up the second order LLM system with Claude, I realized that the way I had it required a Cladude API key, which we don't have access to. I tried to turn it into a locally run Ollama Llama3 server but then very quickly hit the memory limit of the remote server. So I started the process of cloning my repo into my local machine and trying to set it up on my own device.